摘要
针对电气设备热故障定位采用彩色图像融合方法时数据量大的问题,对彩色图像融合以及压缩感知理论进行了研究和归纳,提出了一种基于压缩感知的红外与可见光彩色图像加权融合算法。首先分别将红外热图像与可见光彩色图像分解为单独的3个R、G、B通道,运用小波基对分离出来的各通道数据进行了稀疏表示;其次利用高斯随机矩阵对稀疏数据进行采样得到了测量值,将对应通道的测量值通过两个权重因子进行了融合;最后通过正交匹配追踪重构算法(OMP)对各通道的融合数据进行了重构,将得到的3通道重构数据恢复为融合后的彩色图像。研究结果表明,该算法能用比传统彩色图像融合少50%的采样点来实现融合;在平均梯度、峰值信噪比、空间频率以及信息熵上与传统数据融合指标相当,较好地保留了红外热图中的温升区域信息和可见光背景信息。
Aiming at the problems of large date in color image fusion of the thermal fault location of electrical equipment, researches and in-ductions were made for the color image fusion and compressed sensing theory, a kind of infrared and visible color image fusion algorithm based on compressed sensing was proposed. Firstly, the infrared thermal image and visible color image date was decomposed into three chan-nels of R, G, B, each channel of the date was represented by wavelet-basis. Secondly, Gauss random matrix was used to sample the sparse data and get the measured value, the measurement value of the corresponding channel was fused by the weighting factors. Finally, the orthog-onal matching pursuit reconstruction algorithm （ OMP） was performed to reconstruct the fusion data of each channel, the reconstruction data was restored to the color fused image. The results indicate that the proposed algorithm can achieve fusion with less than 50% of the sample points. The value of AG, PSNR, SF, IE is close to the index of traditional data fusion, which maintaines the information of high temperature area and background area perfectly.
出处
《机电工程》
CAS
2017年第6期674-679,共6页
Journal of Mechanical & Electrical Engineering
基金
上海市高校教师创新基金资助项目(1S10302020)
